Scale-Down of oxygen supply in bioprocess
development with Corynebacterium glutamicum
vorgelegt von
Dipl.-Ing.
Friedrich Käß
geb. in Braunschweig
von der Fakultät III - Prozesswissenschaften
der Technischen Universität Berlin
zur Erlangung des akademischen Grades
Doktor der Ingenieurwissenschaften
- Dr.-Ing. -
genehmigte Dissertation
Promotionsausschuss:
Vorsitzender: Prof. Dr. Roland Lauster, Techn. Universität Berlin
Gutachter: Prof. Dr. Peter Neubauer, Techn. Universität Berlin
Gutachter: Prof. Dr. Marco Oldiges, Forschungsz. Jülich
Gutachter: Prof. Dr. Peter Götz, Beuth Hochschule Berlin
Tag der wissenschaftlichen Aussprache: 18.12.2015
Berlin 2015
List of original articles
The presented results have been published in the scientific journals Bioprocess
and Biosystems Engineering (BPBSE, impact factor 1.8 as of 2013/2014) and
Microbial Cell Factories (MCF, impact factor 4.3 as of 2014). Please refer to
the websites of Springer International Publishing AG (BPBSE) and BioMed
Central Ltd (MCF) to obtain authorized copies.
Paper I1:
Käß, F.; Prasad, A.; Tillack, J.; Moch, M.; Giese, H.; Büchs, J.; Wiechert, W.;
Oldiges, M. Rapid assessment of oxygen transfer impact for
Corynebacterium glutamicum, Bioprocess Biosyst Eng 37, 2567–2577
(2014).
Paper II2:
Käß, F. Junne, S. Neubauer, P. Wiechert, W. & Oldiges, M. Process
inhomogeneity leads to rapid side product turnover in cultivation of
Corynebacterium glutamicum, Microb. Cell Fact. 13, 6 (2014).
Paper III3:
Käß, F.; Hariskos, I.; Michel, A.; Brandt, H.-J.; Spann, R.; Junne, S.;
Wiechert, W.; Neubauer, P.; Oldiges, M. Assessment of robustness against
dissolved oxygen/substrate oscillations for C. glutamicum DM1933 in two-
compartment bioreactor, Bioprocess Biosyst Eng 37, 1151–1162 (2014).
List of presentations at international conferences
GVC/DECHEMA 2013, Bad Wildungen (presentation):“Scale-down study of
oscillations in oxygen and substrate supply for C. glutamicum“, F. Käß, I.
Hariskos, A. Michel, R. Spann, S. Junne, W. Wiechert, M. Oldiges
ACHEMA 2012, Frankfurt (presentation): “Microtiter scale determination
of growth and productivity in Corynebacterium glutamicum”, F. Käß, A.
Prasad, W. Wiechert, M. Oldiges
BioProScale 2012, Berlin (presentation): “Effects of oxygen transfer
limitation and inhomogeneous supply in Corynebacterium glutamicum”, F.
Käß, M. Oldiges
ECAB 2011, Berlin (poster): “Corynebacterium glutamicum: metabolic
impact of changes in oxygen supply determined in microscale cultivation”,
F. Käß, A. Prasad, W. Wiechert, M. Oldiges
Acknowledgements
I thank the students whom I had the privilege to supervise: Ioanna, Arjun and
Robert, whose combined efforts have made this project possible. I thank Hans-
Jürgen Brandt, who designed and constructed the two-compartment reactor in
Jülich and handled pilot scale reference cultivations, Stefan Junne, who
provided know-how and lots of support with the original scale-down reactor in
Berlin, and Peter Rohe, who facilitated the integration of oxygen transfer
screenings into the BioLector®/JUBOS-framework.
Furthermore, I thank my collaborative partners: the bioprocess engineering
group at TU Berlin for support with two-compartment reactor technology and
application, Heiner Giese at the chair of fermentation and bioreactor
technology at RWTH Aachen for conducting micro-RAMOS experiments,
Jana Tillack of the ModSim Group at IBG-1 for developing custom data-
processing, m2p-Labs for reference data in support of oxygen transfer
screening, and the remaining coauthors of my manuscripts who provided
analytical and data-processing supports.
Most of all, I thank my supervisors for the fulfilling and challenging project
that resulted in the presented studies: Marco for his supervision and full-scale
introduction into the academic world of conferences, grants and publications,
and Peter for invaluable structure and planning advice along with organizing
the crucial support of the whole BVT-group.
The project which lead to this dissertation was funded in the Consortium „Corynebacterium:
Improving flexibility and fitness for industrial production“ by the German Bundesministerium
für Bildung und Forschung (grant no. 0315589A). Evonik Industries acted as an industrial
partner, supplying working strain and infrastructure.
Continuation of research on the inhomogeneity scale-down topic is performed in the project
“Systematic consideration of inhomogeneity at the large scale: towards a stringent
development of industrial bioprocesses (SCILS)”, at IBG-1 and TU Berlin, ERA-IB grant.
Table of Contents
1 Abstract ..................................................................................................................................... 1
2 Introduction ............................................................................................................................... 4
2.1 Motivation ......................................................................................................................... 4
2.2 Background: challenges in bioprocess development ......................................................... 6
2.2.1 Common strategies ..................................................................................................... 6
2.2.2 Reaching ideal economy of scale ............................................................................... 7
2.3 Background: metabolism of C. glutamicum .................................................................... 11
2.3.1 Respiration and substrate spectrum .......................................................................... 11
2.3.2 Metabolic stress ........................................................................................................ 12
2.3.3 Production of L-lysine .............................................................................................. 13
2.4 Structure of this work ...................................................................................................... 15
3 Results ..................................................................................................................................... 16
3.1 List of publications with authors’ contributions .............................................................. 16
3.2 Scale-down of oxygen supply assessment into a screening system ................................ 18
3.3 Scale-down of oxygen transfer inhomogeneity into lab-scale ......................................... 26
4 Discussion ............................................................................................................................... 31
4.1 Scale-down of oxygen supply assessment into a screening system ................................ 31
4.1.1 Evaluation of oxygen supply screening method ...................................................... 31
4.1.2 Evaluation of oxygen supply screening results ........................................................ 33
4.2 Scale-Down of oxygen transfer inhomogeneity into lab-scale ........................................ 37
4.2.1 Evaluation of relevance for industrial application ................................................... 37
4.2.2 Evaluation of inhomogeneity scale-down results ..................................................... 40
4.3 Conclusion & outlook ...................................................................................................... 46
4.3.1 Oxygen supply screening ......................................................................................... 46
4.3.2 Scale-down of process inhomogeneity ..................................................................... 47
4.3.3 Outlook ..................................................................................................................... 48
5 Literature ................................................................................................................................. 51
1
1 Abstract
Corynebacterium glutamicum has large-scale industrial applications in the
production of amino acids, e.g. L-lysine, and serves as a platform organism for
new products. Challenges arise because production scales for biological bulk
products often reach reactor working volumes of several hundred cubic meters,
while strain evaluation and process development is based on lab scale
assessments. Oxygen supply is a frequent driver of scaling complications. This
dissertation introduces scale-down methods for oxygen transfer assessments,
and demonstrates them on C. glutamicum ATCC13032 (wildtype) and
DM1933 (L-lysine producer).
The first topic of this work is the development of an oxygen transfer screening
in a down-scaled microtiter plate format. The method is demonstrated by
characterizing the impact of oxygen supply limitation on process yields, side
product secretion, and growth behavior in a batch process. High process
robustness against oxygen supply limitation is observed, and optimal supply
conditions are identified. The second topic is the application of established
two-compartment scale-down setups for analysis of inhomogeneous oxygen
supply in fed-batch environments. The studies demonstrate a remarkable
robustness of C. glutamicum against oscillating oxygen and substrate
availability, and expose some of the mechanisms which underlie the metabolic
flexibility, e.g. intermediate side product formation/reabsorption between
reactor zones. Multi-omics analysis is performed to gain a deeper
understanding of the physiological properties behind inhomogeneity
resistance.
Application of the presented methods has demonstrated that C. glutamicum is
highly adjusted to oxygen transfer limitation in homogeneous and
inhomogeneous process environments, which is an asset for industrial
commodity bioprocess development. Going forward, a model workflow is
proposed to apply the established scale-down methods for other aerobic
2
bioprocess developments, which could improve the resulting process
performance and minimize risks for failure during scale-up.
Kurzzusammenfassung
Corynebacterium glutamicum wird zur industriellen Aminosäureproduktion
eingesetzt, z.B. für L-Lysin, und dient als Platform-Organismus für weitere
Produkte. Bioprozesse können einen Maßstab von mehreren hundert
Kubikmetern erreichen. Den Ausgangspunkt bilden jedoch Versuche im
Labormaßstab, was zu Komplikationen führen kann. Einer der maßgeblichen
Faktoren ist Sauerstoffversorgung. Diese Dissertation führt Methoden zur
Maßstabsverkleinerung für die Untersuchung von Sauerstoffversorgung ein,
und demonstriert sie an den Stämmen C. glutamicum ATCC13032 (Wildtyp)
und DM1933 (L-Lysin Produzent).
Das erste Element dieser Arbeit ist die Entwicklung eines Sauerstofftransfer-
Screenings im maßstabsverkleinerten Mikrotiterplattenformat. Es erfolgt eine
Charakterisierung des Einflusses von Sauerstofflimitation auf
Prozessausbeuten, Nebenproduktbildung und Wachstumsverhalten im Batch-
Prozess. Eine hohe Prozess-Robustheit gegen Sauerstofflimitation wird
festgestellt, und optimale Versorgungsbedingungen werden identifiziert. Als
zweites Thema werden etablierte Zweikompartiment-Reaktoren für
Maßstabsverkleinerung inhomogener Sauerstoff- und Substratversorgung in
Fed-Batch Prozessen angewendet. Im Ergebnis zeigt C. glutamicum eine hohe
Robustheit bei oszillierender Sauerstoff- und Substratverfügbarkeit.
Zugrundeliegende Mechanismen werden offengelegt, z.B. Bildung und
Wiederaufnahme von Nebenprodukten in einzelnen Reaktorzonen. Multi-
Omics Analysen liefern dazu tiefere Einblicke in die Grundlagen dieser
Robustheit.
3
Die präsentierten Methoden zeigen, dass C. glutamicum stark an
Sauerstofflimitation und inhomogene Prozessbedingungen angepasst ist. Dies
stellt einen Vorteil für die Anwendung in Bioprozessen von großem Maßstab
dar. Für die Zukunft wird ein Modellprozess zur Verbesserung aerober
Bioprozessentwicklung vorgeschlagen, der Maßstabsverkleinerungen
einschließt. Dies kann zur Verbesserung der Performance neuentwickelter
Bioprozesse führen, und die Risiken für Fehlschläge bei der
Maßstabsvergrößerung minimieren.
4
2 Introduction
2.1 Motivation
Industrial bioprocess development gets more important with every new
demand for a chemical, pharmaceutical or nutrition product that arises within
the world market. One of many driving factors is that industrial biotechnology
needs to replace existing petrochemical process chains, because regenerative
sources will eventually be without alternative. Higher biocompatibility of the
substrates is another of many reasons why commodity chemicals from
biotechnological origin are often preferred- not only in the field of food, feed
and pharmaceuticals. However, bioprocess development has some unique risks
which have to be mitigated in order to create and maintain reliable process
chains.
One challenge is that using biological entities for production requires ambient,
controlled reaction environments, because correct operation of the cell
factory’s enzymatic machinery is bound to physical and chemical constraints.
Furthermore, carbon fluxes along metabolic networks can easily be redirected
if a necessary compound gets limiting or regulation is triggered, so that futile
reactions reduce the product yield if there is an environmental disturbance.
Therefore, bioprocesses are hard to optimize. It is necessary to find the best
strategies for identification of ideal bioprocess conditions, and to match the
biological demands with technical feasibility during process development.
One of the most critical parameters for many bioprocesses is oxygen supply,
because the low solubility and high metabolic uptake of molecular oxygen can
make precise in-process-control very demanding. Failing bioprocess
development is often caused by neglected oxygen supply requirements during
research, development, and transfer to final process scale. This thesis is an
approach for comprehensive consideration of oxygen supply requirements
during bioprocess development, and aims to supply the necessary tools to deal
with its impact in a typical aerobic development chain. The organism which is
5
assessed, C. glutamicum, and the established techniques should serve as an
example how to improve aerobic bioprocess development. The discussion
aims to point out the scope, validity, limitations, and combined added value
for the presented methods. It also aims to point out the potential for transfer of
the practical conclusions to new studies which consider oxygen supply impact
during bioprocess development.
6
2.2 Background: challenges in bioprocess development
Typical strategies that are applied in the development of bioprocesses should
be considered as a context for the presented results.
2.2.1 Common strategies
Biological systems express complex and highly regulated behaviors, which are
activated by multiple external triggers. In the ideal case, biological systems for
industrial application are adjusted to technically feasible bioreactor conditions,
which can be one of the targets for rational strain design. However, due to
biological complexity there is always the potential for unwanted metabolic
reactions, which can have negative impact on growth or productivity in a
bioprocess. Avoiding unwanted reactions means avoiding the respective
trigger conditions. Therefore, physical and chemical control of reaction
environments is necessary, and suitable parameter ranges must be defined for
every bioprocess. Process development can be regarded as the pursuit of fixed,
suitably characterized process parameter ranges, which should be determined
as easily and quickly as possible. This ultimately results in a process validation
with a subsequent control strategy.
A major complication for designing optimal processes is that bioprocess
environments often change throughout the development phases. System
properties like process scale or reactor type must be adjusted based on
economic constraints, leading to repeated transfers between development
stages. This means that not all environmental process parameters can be
maintained as constants, even if sophisticated transfer strategies are applied.
Depending on the robustness of the biological entity, any of these changes may
pose a threat to process performance. Industrial biotechnology has devised
strategies how to deal with this development challenge. The applied strategy is
set according to the target application.
7
High-margin/ high-regulation fields, e.g. the pharmaceutical industry, try to
keep the degree of process changes after initial development as low as
possible. The intention is to minimize the risk of performance loss, while
maximizing the development speed for earliest possible process validation.
This usually happens at the expense of process efficiency, because
optimization is limited to the early research phase, and scaling effects are
neglected. A corresponding method to this strategy is to apply numbering-up
instead of scaling-up4, i.e. increasing the number of small-scale production
units instead of increasing the production unit size. The roller bottle
technology for mammalian cell culture5 is a typical example for this strategy,
which has a high prevalence even though cell carrier technologies have long
been established.
In contrast, low-margin/ medium regulation fields, e.g. commodity chemicals
and feed products, aim to reach the ideal economy of scale. The main risk is
that the process may not be competitive in the final reactor system, which
means that the price of substrates and energy may exceed the profit margin of
the product. Therefore, maximal process efficiency is targeted through
repeated optimization steps during development, e.g. during the transfer from
shake-flasks in the initial development stage to stirred tanks of several hundred
cubic meters in the commercial stage. This requires repeated bioprocess
transfers with multiple adaptations in the reaction environments, and
constitutes one of the most challenging tasks in bioprocess engineering.
2.2.2 Reaching ideal economy of scale
The main challenges of large-scale, low-margin bioprocess development are
the transfer from initial process-related data acquisition (e.g. picking the most
promising mutant strain from a screening) to lab scale process development,
and the transfer from lab to production scale. These steps introduce various
changes to the bioprocess which might act as harmful triggers for biological
regulation. They are therefore often separated into multiple steps with separate
optimization cycles. In the following, pitfalls for this strategy from the process
8
development perspective are highlighted. The chosen example corresponds to
a large-scale fermentation optimization with a product-secreting
microorganism, which is a typical challenge faced in industrial amino acid
production (e.g. L-lysine). Development works in initial strain design and
downstream treatment are assumed to be the starting and end point of
development, respectively, and are therefore not part of the assessment:
In the beginning of a development chain, a number of potential production
strain candidates have been generated that show promising metabolic traits.
Strains may have been isolated by selection for successful target gene transfer
(e.g. antibiotic resistance), or by more sophisticated approaches, for example
product-specific induction of fluorescence with intracellular biosensors 6. This
is followed by a screening procedure in order to select the most suitable
strains.
Screening is typically understood as the process of comparing large pools of
strains or cultivation parameters under simplified process conditions. The
target of screening is to narrow down the options for subsequent bioprocess
optimization, which takes up more resources and should only be performed
with the most promising strains or conditions. Systems for screening are
usually small-scaled, e.g. shaken microtiter cultivation or shake flasks, due to
the high throughput of strain generation procedures and vast parameter spaces
like media composition. Screenings are performed based on a limited set of
indirect performance parameters, e.g. fluorescence or sufficient viability to
indicate target gene transfer. More sophisticated approaches may provide first
quantitative performance information, e.g. product yield per substrate. Ideally,
the screening conditions anticipate elements of the later production scale, e.g.
industrially relevant media preparations or culture times. Eventually, the most
promising strain candidates or conditions are identified. This is the usual
output of a screening stage, which is then used as a basis to proceed with
process optimization.
After screenings, potential production strain candidates have to be compared
based on their performance under feasible target process conditions. Typical
9
biological traits for comparison can be productivity, viability, robustness
against long culture times, susceptibility to contaminants, or other general
criteria for the desired route of further process development. Several traits may
also be combined into more complex target criteria, e.g. space-time-yield in a
target reactor system. Process optimization requires a controlled process
environment, e.g. a stirred tank reactor. Also, the reactor system will typically
show similarity to commercial scale reactors, keeping in mind that the
subsequent steps of process development should be kept as simple as possible.
This means that the strain candidates are compared in a high number of
typically Design-of-Experiments (DoE)-based stirred-tank reactor cultivations
with the respective analytics. The result will be a balanced ranking of the
producer strain candidates under optimal conditions, and facilitate
identification of the best-suited strain to proceed in the process development
chain.
At the end of the lab-scale optimizations, the fixed bioprocess with best-
performing strain and optimized process conditions will be subjected to
incremental increases in process scale. These scale-up steps are usually
twofold to tenfold, depending on the expected robustness of the bioprocess.
Scale-up criteria are selected based on the available systematic knowledge
about the biological entity. Example criteria can be similarity of dissolved
oxygen or power input per volume in case of critical oxygen dependency. The
aim of the scale-up is to maintain the desired productivity through repeated
optimization. Necessary adaptations in cultivation parameters are typically
made whenever a loss of productivity is observed, or when technical or
commercial restrictions cannot be avoided.
Application of the described process development strategy means that the
target of the development phases changes in three stages. Initial screenings or
preliminary studies aim to pick only the most promising strains or parameter
settings from a high amount of candidates. Lab-scale development optimizes
the process as far as possible. Subsequent scale-up aims to maintain
productivity from the optimized process over repeated transfer steps. The
10
advantage of this development strategy is that only high-performing
bioprocesses can reach the final reactor stage, because all others will be
rejected in previous stages. The disadvantage is that the changing targets of the
development phases can lead to a disconnection between the initial strain
selection and the commercial scale process requirements, so that ideal strains
or conditions are wrongly discarded before the final development stage. For
example, this could mean that a strain that has ideal robustness against
unavoidable stress factors of the industrial scale never reaches large reactors,
because it is discarded when inferior product yield is observed under screening
conditions. A strain like this might be the optimal candidate for the final
reactor scale, but would never pass initial stages of process development.
Therefore, it is highly advisable to refine the typical development strategy with
more consistent approaches that are aimed at the final commercial production
conditions. This thesis is an approach for such a refinement (see 2.4).
11
2.3 Background: metabolism of C. glutamicum
C. glutamicum is a highly characterized industrial microorganism with a broad
range of applied production strains for many commodity products, which is
most thoroughly documented in the recent “Handbook of Corynebacterium
glutamicum”7. In the following, several relevant properties for the presented
studies are summarized.
2.3.1 Respiration and substrate spectrum
Metabolic properties of C. glutamicum and other members of the genus
Corynebacterium have been investigated extensively in the past due to their
relevance in industry and medicine. This comprises a broad knowledge about
the respiratory chain, which was summed up by Bott and Niebisch8. There is a
broad spectrum of substrates which can be metabolized9, although the
industrially relevant media preparations are mainly based on sucrose from
cane or starch hydrolysates as carbon source7.
C. glutamicum is an aerobic respiratory organism. It utilizes both substrate
level phosphorylation and oxidative phosphorylation as a source of energy for
growth and specific biosyntheses. Oxidative phosphorylation provides the
surplus of metabolic energy and is therefore crucial for growth. There is also
anaerobic or micro-aerobic potential for biotransformations, in which
C. glutamicum can be used as a catalyst. However, this is a less common
application, because the growth and self-maintenance of the organism is one of
the most attractive features for large-scale industrial application. This means
that availability of oxygen is a prerequisite for sustainable cultures that rely on
active growth or maintenance metabolism. Concerning oxygen metabolism,
there is a high degree of flexibility in the branched breathing chain of
C. glutamicum, which likely provides options for adjusting to different levels
of oxygen availability. The fundamentals of these oxygen-dependent
properties of C. glutamicum have been summed up in a publication of Bott and
Niebisch in 20038.
12
There is a wide range of substrates which can be oxidized by C. glutamicum in
order to facilitate electron transfer to reducing equivalents, which facilitates
the maintenance of metabolic function in many biochemical environments.
Several of these substrates are also known side-products of the organism,
which get secreted – and in some cases reabsorbed - most pronouncedly under
non-ideal growth conditions. The most prominent example of side product
secretion occurs under oxygen supply limitation, when organic acids are
secreted as side products, e.g. described for L-lysine production by Ensari and
Lim10. The substrate spectrum of C. glutamicum also contains a much wider
scope of sugars, acids and alcohols, which is facilitated by a comparatively
versatile set of enzymes for uptake and breakdown of these compounds7. The
combination of a broad substrate scope and versatile oxygen-dependent
metabolism allows a high flexibility of application in bioprocesses. In order to
avoid energy loss through side product generation, most industrial applications
of C. glutamicum are centered on fed-batch processes with limited carbon
source. At homogeneous conditions, this ensures that metabolic activity is
limited, so that oxygen supply limitation and side product generation can be
avoided.
2.3.2 Metabolic stress
There are several kinds of stress which are relevant for industrial
microorganisms, most prominently in large-scale process environments. These
can get particularly critical in inhomogeneous cultivation systems, where
individual cells face rapidly changing, i.e. oscillating stress conditions. The
most process-relevant stress factors for microorganisms with a highly active
metabolism are hyperosmotic stress and oxygen or substrate supply limitation,
because these are directly related to feed zone conditions in industrial
substrate-limited environments. C. glutamicum production strains are
comparatively well-adjusted to osmotic stress in a reasonable range, as was
characterized for an L-lysine process by Guillouet et al.11. On the other hand,
oxygen and substrate supply limitation has direct impact on the substrate
13
conversion and biomass yield, i.e. increases the energy demand for
maintenance metabolism7.
There are also other kinds of metabolic stress, e.g. shear stress or pressure
gradients, against which C. glutamicum does not seem to be susceptible in a
bioprocess-relevant range. Acidity of medium can also play a role: Follmann
et al. report that growth is negatively influenced below pH 612. Carbon dioxide
saturation can be considered as another potential stress factor, because some
metabolic reactions rely on sufficient availability, while elevated levels of CO2
can also have other regulation effects13. Except for oxygen and substrate
limitation, all these stress factors can be considered as relatively unproblematic
in typical production conditions, which are characterized by controlled
reaction environments in actively agitated reactors.
2.3.3 Production of L-lysine
The current annual production of L-lysine by mutants of C. glutamicum is in
the range of 2.2 million tons, which makes it one of the leading products
generated by biotechnological means14. L-lysine is an essential amino acid for
feed applications, and L-lysine producing C. glutamicum strains have been
established as early as 195615,16 with continuous improvements in strains and
processes ever since then. The high competitiveness of the organism for amino
acid production has to do with its secretion capacity for amino acids, as well as
highly efficient central metabolic pathways. L-lysine is synthesized from the
tricarboxylic acid (TCA) cycle intermediate product oxaloacetate and various
precursors (NADPH+H+, pyruvate, glutamate/NH4+) through a sequence of
enzymatic reactions, starting with L-aspartate7. Robust strains have been
constructed with various modifications, one of which is the non-auxotrophic
C. glutamicum DM193317 that is applied in this work. The modifications of
this strain comprise several deletions, duplications and new insertions which
facilitate overproduction and secretion of L-lysine (Δpck pyc(P458S)
hom(V59A), 2 copies of lysC(T311I), 2 copies of asd, 2 copies of dapA, 2
copies of dapB, 2 copies of ddh, 2 copies of lysA, 2 copies of lysE derived
14
from Wildtype C. glutamicum). The strain has comparatively robust growth
and production behavior and can be regarded as a model system for currently
applied industrial mutants.
15
2.4 Structure of this work
Applying a scale-down approach means decreasing the size of a system for
gaining a benefit over the original scale. In process engineering, this usually
has the target to save time and resources by obtaining process performance
data without having to run full-scale experiments. This thesis introduces two
scale-down elements for oxygen supply assessment into bioprocess
development. Both are performed with C. glutamicum ATCC13032 (wildtype)
as a proof-of-concept for the method itself, and with the L-lysine producer
C. glutamicum DM1933 as an application example for an industrially relevant
process.
First, oxygen transfer assessment is implemented into a screening workflow
for increased throughput1. This facilitates more thorough assessment of the
oxygen supply impact as a basis for further process development. This new
scale-down methodology for a process parameter assessment is discussed in
section 4.1. Second, the inhomogeneity of large-scale oxygen supply is
brought to the lab-scale in two compartment reactors. Two manuscripts
contain a general assessment of the wildtype strain2 and a more detailed
analysis of the L-lysine process3. This provides more information about the
process robustness and scalability. The inhomogeneity scale-down is discussed
in section 4.2.
Subsequently, the combined scope of results is discussed in section 4.3. The
conclusion for integration of the presented scale-down elements into
bioprocess development workflows is drawn. As an outlook, the section
contains a proposed model workflow for consistent consideration of oxygen
supply requirements throughout the development phases.
16
3 Results
3.1 List of publications with authors’ contributions
Paper I1: Rapid assessment of oxygen transfer impact for C. glutamicum
Käß F, Prasad A, Tillack J, Moch M, Giese H, Büchs J, Wiechert W, Oldiges
M.
Authors’ contributions: FK designed the experiments, developed and
validated the experimental methods and prepared the manuscript. AP and FK
performed and evaluated the experiments. JT developed the data processing
for volume-dependent calibration in MATLAB. MM and HG conducted
analytical process development and performed tests. JB is the scientific
supervisor in development of the applied RAMOS analytics. MO and WW
initiated the project. MO is the principal investigator, supported with
conception and manuscript preparation. All authors read and approved the
final manuscript.
Paper II2: Process inhomogeneity leads to rapid side product turnover in
cultivation of Corynebacterium glutamicum
Käß F, Junne S, Neubauer P, Wiechert W, Oldiges M.
Authors’ contributions: FK designed the experiments, developed and
validated the experimental methods and prepared the manuscript. FK and SJ
performed and evaluated the experiments. PN and SJ provided the previously
published two-compartment reactor setup, helped with conception of
experimental methods, and participated in conduction of experiments. MO and
WW initiated the project. MO is the principal investigator, supported with
conception and manuscript preparation. All authors read and approved the
final manuscript.
17
Paper III3: Assessment of robustness against dissolved oxygen/substrate
oscillations for C. glutamicum DM1933 in two-compartment bioreactor
Käß F, Hariskos I, Michel A, Brandt HJ, Spann R, Junne S, Wiechert W,
Neubauer P, Oldiges M.
Authors’ contributions: FK designed the experiments, developed and
validated the experimental methods and prepared the manuscript. IH, RS and
FK performed and evaluated the experiments. AM conducted the
transcriptome analysis. HJB designed and built the two-compartment reactor
which was used for most cultivations. PN and SJ provided the previously
published two-compartment reactor setup for reference, which the described
reactor is based on, helped with conception of experimental methods, and
participated in conduction of experiments. MO and WW initiated the project.
MO is the principal investigator, supported with conception and manuscript
preparation. All authors read and approved the final manuscript.
18
3.2 Scale-down of oxygen supply assessment into a screening
system
Method development
Paper I1 starts with an assessment of the importance of oxygen supply in
bioprocess development. It is emphasized that each commercial bioprocess
development step has unique requirements and characteristics in oxygen
supply. Shaken bioreactor systems form the typical setup of early strain and
process characterization. The rationale of using the BioLector® as a
development platform for fast oxygen supply assessment is presented as a
combination of analytic versatility and easy bioprocess scale-up. Also, the
system has a broad range of oxygen transfer capacities that can be reached,
with a very high peak value of over 100 mmol·L-1·h-1 that spans the feasible
transfer rates for stirred tank cultivation. The high degree of parallelization in
bioprocess conduction and analysis means that scale-down of bioprocess lab
assessments into the BioLector® is a promising approach. The target of the
development in the presented study is therefore to set up and test the
assessment of oxygen supply impact in the BioLector® system by
characterizing oxyen supply impact on C. glutamicum as a model organism for
industrially relevant conditions. This should provide speed, data abundance
and easy scale-up potential. Main challenges are identified in the necessary
online analytics that have to be adjusted to said purpose.
The functionality of the oxygen transfer screening approach is detailed in
Supplementary Fig. 61. Utilized elements are the regulation capacity of
headspace oxygen content in the BioLector® and the availability of
unpublished characterization data for flowerplate cultivation geometries at
various filling volumes (sulfite oxidation-based determination of maximum
oxygen transfer capacity (OTRmax), m2p-Labs Baesweiler). By using
HENRY’s law. the oxygen transfer capacity is adjusted with a combination of
headspace oxygen control and filling volume changes.
19
Experimental design was performed as a pattern of batch cultivations that
spans feasible oxygen supply ranges. Cultivation conditions were set as a
batch process at fixed agitation. Samples were prepared by a liquid handling
station, which improves reproducibility due to standardized sample preparation
times (Methods section1). Replicate samples for offline measurement were
pooled to achieve sufficient volume for HPLC-based analytics of side
products. Reference cultivations in a MicroTiter Plate Respiration Activity
Monitoring System (MTP-RAMOS) were performed at identical cultivation
conditions as in the BioLector® (flowerplate geometry, agitation type, culture
inoculation density and media1).
Data processing & method verification
The main challenge for the study was identified in the analytical codependency
for online analytics of C. glutamicum cultures (see Method section and
Supplementary Fig. 71). The cultivation at changing filling volume is not a
typical application form of the BioLector®, which explains why this unusual
analytical behavior has not been reported previously. Risk mitigation is
presented as a MATLAB calibration procedure for backscatter-based biomass
determination. Development of this technique was achieved by non-linear
interpolation of the parameter space biomass - filling volume- backscatter
from reference measurements. The underlying plots of the interpolated
function are depicted in Fig.11. Stationary biomass samples were measured in
the BioLector® as reference for the interpolation. Accuracy of the biomass
determination was evaluated as sufficient for a screening application, although
very low biomass and filling volume carry an increased risk of measurement
inaccuracy (Fig.11). It is speculated that this may be caused by increased
surface reflection at low turbidity. As a result, cultivation experiments in the
conducted screening were performed with multiple replicates to minimize
measurement errors.
Method verification was performed by comparing oxygen-limited BioLector®
cultivations to similar process conditions of filling volume and headspace
oxygen content in MTP-RAMOS. Both systems returned similar oxygen
20
supply limitation times when compared on the basis of OTR (MTP-RAMOS)
and DO (BioLector®) (Fig. 2B1).. Oxygen-limited growth behavior was
observed in clear dependence on the volume-adjusted maximum oxygen
transfer capacity (OTRmax) (Fig. 2A1). Distinct peak values of oxygen transfer
capacity could be identified for each assessed filling volume. The observed
values for OTRmax in MTP-RAMOS suggest a systematic deviation to
BioLector® characterization data, which were obtained by sulfite oxidation
reference experiments (Fig. 2A1).
It is concluded that oxygen-limited cultivation of C. glutamicum in the
BioLector® yields characteristic plateau-shaped oxygen transfer conditions
with distinct OTRmax for each filling volume, which is the crucial requirement
for conduction of the oxygen transfer screening.
Metabolic analyses
The screening was performed by scattering cultivations with distinct OTRmax
in the range from 0.8 to 35 mmol·L-1·h-1. Application of highly limited oxygen
transfer conditions showed the dependency of growth rate and biomass yield
on OTRmax, which is clearly demonstrated by growth curve comparison in a
single flowerplate with variation in filling volumes for C. glutamicum
wildtype (Fig.3A and 3B1). The growth curves follow an approximately linear
profile after initial batch growth, meaning that there is a decreasing growth
rate during the linear phase, and that biomass increase is roughly proportional
to the oxygen that is supplied to the culture. Consequently, lower OTRmax
leads to extended linear growth phases, and to increased process time until the
stationary phase is reached. Optode data facilitates the identification of the
relationship between growth and oxygen supply limitation (Fig. 3A1), which
shows a delay of several hours between the full consumption of dissolved
oxygen (OTR = OTRmax) and the change of growth characteristics from a near
exponential to a linear profile. Sampling of the culture supernatant at
synchronized timepoints shows the increasing degree of organic acid side
product secretion with oxygen-limited growth (Fig. 3C1). The identified side
products lactate, succinate and acetate are formed in a seemingly conserved
21
sequence and later reabsorbed when carbon source becomes limiting (Fig.
3D1).
Secretion of the side products is necessary to maintain redox balance in an
environment of oxygen limitation, where NAD+ is regenerated through
enzymes as lactate dehydrogenase instead of oxidative phosphorylation. Later
reabsorption of organic acids as carbon source means that lower quality
substrate is used for growth, which in turn can lead to irregular growth
patterns, which is observed for very low oxygen transfer capacities. In
summary, metabolic energy is clearly reduced by the lack of oxidative
phosphorylation, which is assumed to be the driver of decreased growth at
oxygen supply limitation. The biomass yields can be plotted over OTRmax ,
which gives the most striking visualization of the detrimental effect of oxygen
supply limitation for C. glutamicum wildtype (Figure 1, not shown in
publication). Biomass yield roughly follows a saturation curve, meaning that
oxygen supply limitation has an increasingly detrimental effect below a
threshold value of ca. 12 mmol·L-1·h-1 for the wildtype strain.
22
Figure 1: biomass yield on substrate for C. glutamicum ATCC13032 (wildtype) as a function
of maximum oxygen transfer in batch process, as determined in newly developed oxygen
transfer screening procedure (BioLector®)
After analysis of wildtype behavior, the screening was applied to generate a
broad set of process performance data for the L-lysine producer strain
C. glutamicum DM1933 (Fig. 4A, Fig. 4B, Fig. 5A and Fig. 5B1). One result
was that the optimum OTRmax lies above 22.8 mmol·L-1·h-1, which results in
batch growth kinetics that are unaffected by phases of oxygen transfer
limitation (Fig. 4A1). In contrast, the sufficient OTRmax for maximum biomass
yield and product yield is in the range of 15 mmol·L-1·h-1, meaning that the
intermediate range has a slowing effect on growth without loss of yield. The
relationship between OTRmax and process yield is consistent for both biomass
and product (Fig. 5A and Fig. 5B1). Biomass yield roughly follows a
saturation curve, meaning that oxygen supply limitation has an increasingly
detrimental effect below the threshold value of 15 mmol·L-1·h-1. Product yield
is largely unaffected by OTRmax. An exception to this behavior can be found in
the product formation at very low oxygen transfer, where a slight increase in
L-lysine formation was observed. This growth decoupling of product
formation is in accordance with findings of Neuner et al.18. In this case, it
demonstrates how a performance screening may serve to identify new
23
bioprocess working points and optimization potential. However, it should be
noted that the respective product yield was obtained after a highly prolonged
culture time, because low OTR lead to a six fold higher duration until full
carbon consumption. It is therefore unlikely that the newly identified L-lysine
production optimum is a desirable target for large-scale application due to
drastically reduced space-time yield.
Scale-Up into lab-scale stirred tank
Some of the cultivations that were performed in Paper I1 were also assessed in
a laboratory scale (5 L) stirred tank reactor. The target was a direct scale-up of
OTRmax conditions in order to verify if process performance could be
reproduced. The scale-up was challenged by the identified variation in OTRmax
data between sulfite oxidation (BioLector® characterization data) and balance-
based OTRmax (RAMOS), because the most informative process performance
effects that were identified in the screening occurred at extremely low oxygen
supply. These conditions could not be verified in MTP-RAMOS before scale-
up because of time and resource constraints. Therefore, the scale-up was based
exclusively on the sulfite oxidation reference data which were used as a basis
for the screening. Low oxygen transfer in the stirred tank was achieved
through lowering the gassing rate and stirring speed. Figure 2 and Figure 3
show that although a qualitatively similar behavior was observed concerning
product yields at variation of OTRmax, there is still a high deviation between
the actual process yields.
24
Figure 2: 5 L stirred-tank cultivation biomass yield on substrate (large circles) for C. glutamicum
ATCC13032 (wildtype) as a function of maximum oxygen transfer capacity, comparison to
data from newly developed oxygen transfer screening procedure (BioLector, see Figure 1)
Figure 3: 5 L stirred-tank cultivation biomass yield on substrate (blue circles) and product yield on
biomass (yellow circles) for C. glutamicum DM1933 (L-lysine producer) as a function of
maximum oxygen transfer capacity, comparison to data from newly developed oxygen transfer
screening procedure (BioLector, see Fig. 51)
It can be expected that the main influencing factor that causes the deviation in
yields is indeed the method of OTRmax determination. The systematic
25
underestimation of OTRmax from balance-based approaches in comparison to
sulfite oxidation has been demonstrated in Fig. 21. This phenomenon may have
an even higher impact in cultivations with very low OTRmax, which would
mean that limitation in the BioLector® is less strong than OTRmax data from
sulfite oxidation would suggest. This would explain the quantitatively
different, yet qualitatively similar behavior that was observed in the
uncorrected OTRmax-based comparison of BioLector® and stirred tank reactor.
26
3.3 Scale-down of oxygen transfer inhomogeneity into lab-
scale
Method development
Paper II2 and Paper III3 introduce C. glutamicum cultivation in two-
compartment reactors with stirred tank and plug flow reactor elements (STR-
PFR) for analysis of process inhomogeneity in oxygen and substrate supply.
Fed-batch is the cultivation mode of choice because of its high industrial
relevance and susceptibility to mixing challenges. Paper II2 is the first
published application of STR-PFR reactors for C. glutamicum wildtype. Paper
III3 is a focused metabolic robustness assessment of the L-lysine producer
strain C. glutamicum DM1933. Both studies share their general approach, in
which STR-PFR cultivation at variable plug flow residence time is compared
to homogeneous reference cultivations (stirred tank reactor). Oxygen supply
limitation is achieved in the unaerated plug flow elements due to high
metabolic turnover rates, whereas the stirred tank compartment is kept at
aerobic conditions with sufficient dissolved oxygen at all times. Paper II2 has
the focus of characterizing basic metabolic phenomena that occur under
inhomogeneous oxygen/substrate supply in a batch/fed-batch environment of
C. glutamicum wildtype. It has a broad process scope that takes into account
batch and fed-batch phases with a smaller emphasis of process yields. In
contrast, Paper III3 is a focused assessment of metabolic robustness that
applies multi-omics analyses in an exponential fed-batch process. It is
designed for targeted comparison of various process characteristics with the L-
lysine producing C. glutamicum DM1933 in multiple replicates, including
homogeneous reference cultivation in pilot scale. Process yields and
reproducibility are the central elements of this second study, which applies
highly extended exposure to oxygen supply limitation in the plug flow element
for robustness assessment.
27
Reactor characterization
Defined process inhomogeneity was achieved by using two-compartment
reactor systems with characterized backmixing behavior of the plug flow
elements. Extrusion experiments2 and tracer pulse-based determination of
residence time distribution3 were used to determine the mean residence time at
defined pump circulation settings, and to verify plug flow behavior. The
reactor for the wildtype study (Fig. 12) is a specialized setup with the potential
for gassing the plug flow element through static mixers, while the newly
constructed unit for the L-lysine producer study (Fig.13) was designed with the
target of generating extended residence times in the plug flow compartment.
Paper III3 has an increased degree of biological process characterization, with
studies on the speed of dissolved oxygen depletion (Fig. 2B3) and required
biomass for reaching oxygen supply limitation (Results section3).
PFR-monitoring: immediate responses to step change conditions
Paper II2 is based on a process with two phases: the batch and fed-batch phase
(Fig.22). In both phases, oxygen supply depletion leads to immediate
limitation, as was demonstrated by the lack of dissolved oxygen and the
reduced oxygen uptake in the plug flow compartment (Fig.32). A concomitant
lactate secretion was identified (Fig.4B2) that can be traced along the plug flow
reactor, which is accompanied by a moderate pH drop (Fig.4A2). The extent of
carbon fraction turnover during the fed-batch phase reaches very high extents,
with the majority of glucose-carbon being converted into extracellular lactate
within an oxygen-limited residence time of ca. 87 seconds (Fig.52). The
shorter plug flow residence time of 45 seconds has similar effects to a weaker
extent (Fig.5A2). Glucose uptake of the wildtype was increased compared to
aerobic conditions (Results section2). Determination of pool sizes for
intracellular adenosine phosphates (AMP, ADP, ATP) and redox cofactors
(NAD(P)+/NAD(P)H+H+) did not show any immediate response to oxygen
supply limitation (Fig.62). It is concluded that the regeneration of NAD+
through lactate formation and excretion is the balancing factor that allows
C. glutamicum wildtype to rapidly prevent any redox imbalance that might
28
occur under oxygen supply limitation. This way, substrate uptake and
conversion are not negatively affected by changes in the redox levels (e.g. no
inhibition of glyceraldehyde-3-phosphate dehydrogenase through unfavorable
NAD+/NADH+H+ ratio).
PaperIII3 contains a more drastic challenge of robustness against
inhomogeneity of oxygen and substrate supply for the L-lysine producer strain.
Acidification and side product secretion were observed at oxygen supply
limitation, with the identified side products lactate, succinate and pyruvate
(Fig.53). Pyruvate was the only side product that did not follow the rapid
secretion and reabsorption behavior, which had previously been observed for
lactate in the wildtype study2. Although the residence time under oxygen
supply limitation was extended to ca. three minutes, the extent of carbon
fraction turnover from glucose into side products was around 25% maximum.
Part of this effect may have been caused by lower biomass concentration in the
L-lysine producer study, which decreases the speed at which dissolved oxygen
becomes limiting. Acidification (Fig. 43) was observed to be roughly
proportional to biomass and feed amount. Substrate uptake was identified to
continue at oxygen supply limitation, with a comparable rate to the maximum
uptake capacity (see below). This indicates the robustness of the producer
strain. Metabolic consequences of substrate excess/ oxygen supply depletion
could be identified as an increase in cellular energy charge (Fig. 73), although
the robustness identified in redox cofactors of the wildtype study was also
observed for DM1933.
Culture monitoring: robustness against process inhomogeneity
Paper II2 does not identify changes in the growth behavior of C. glutamicum
wildtype in response to plug flow residence times of up to 87 seconds,
although the database for comparison is limited due to the small amount of
cultivation replicates. Net respiration rates remained similar at homogeneous
reference and inhomogeneous STR-PFR cultivation during the fed-batch phase
(Fig.32), which demonstrates a redistribution of oxygen uptake from limited to
the unlimited zones of the reactor. In the batch phase, however, the oxygen
29
uptake rate of the inhomogeneous culture is depleted by roughly the volume
fraction of the PFR in the two compartment system, which demonstrates that
partial oxygen supply limitation cannot be compensated under batch
conditions (Fig.32). In spite of pH effects during oxygen supply limitation
(Fig.42), base requirement for the cultivations was not increased, which
suggests reversibility of the acidification along with the metabolization of
organic acids in aerobic zones. No net accumulation of organic acids was
observed, which means that the inhomogeneous process is in fact a mixed
substrate situation of primary carbon source and the intermediately formed
side product lactate. In conclusion, the fed-batch mode is assumed to be a
superior working condition for processes with C. glutamicum that are subject
to process inhomogeneity of oxygen and substrate supply. It is deduced that
the microbial organism has a native adaptation to rapidly changing
microenvironments, because the reversible switch to fermentative metabolism
does not impair growth.
Paper III3 was performed with exponential fed-batch, and therefore has a very
reproducible and stable process configuration that is aimed at identifying
changes in metabolic maintenance demands and process yield changes. It does
not contain a comparison to batch phase conditions. Nevertheless, the zonal
redistribution of oxygen uptake in the fed-batch is clearly observable for a full
cultivation period in Fig.33. As in the wildtype study, the producer strain is
capable of compensating partial oxygen supply limitation through increased
breathing activity in the aerobic bulk zone. Targeted analysis of the maximum
substrate uptake capacity (Fig. 63) shows that oxygen/substrate inhomogeneity
does not affect the functionality of the cell’s substrate uptake system and
metabolization speed. Instead, a maximum substrate uptake capacity of ca.
0.95 g·g-1·h-1 is maintained throughout the various degrees of inhomogeneity
that were assessed in the study. This is an indication for similarity in enzyme
availability and regulation. In a multi-omics approach, a set point after
extended exposure to inhomogeneous cultivation conditions was compared to
homogeneous reference cultures in a targeted proteome- and transcriptome-
assessment for enzymes of the central metabolic pathways (Fig. 83).
30
Regulation in response to process inhomogeneity could not be detected for any
of the targeted enzymes. This shows that the reversible switch to fermentative
pathways is exclusively mediated on the enzyme activity level (e.g. allosteric
control, enzyme activities), instead of requiring specific regulation responses
on transcriptome or proteome level. Finally, the strongest robustness indicator
is sustained by the overall similar process yields for L-lysine secretion and
biomass growth for the various inhomogeneity degrees that were assessed in
the study (Fig. 93).
31
4 Discussion
4.1 Scale-down of oxygen supply assessment into a screening
system
Screening for oxygen supply demands during early process development
stages allows characterization of ideal net supply and robustness against
limitation. This constitutes a scale-down for a parameter assessment that
would otherwise be performed at later process development stages with
significantly higher resource constraints. As outlined in the results section, the
methodology in Paper I1 offers an improvement for screening setups whenever
the degree of oxygen supply is a critical parameter for process efficiency. In
addition to the discussion of the results in the original manuscript, there are
two elements that should be elaborated further: the technical quality and
limitations of the screening method itself, and the quality and limitations of the
metabolic dataset that was generated for bioprocesses with C. glutamicum.
4.1.1 Evaluation of oxygen supply screening method
Two key assets are the basis of the presented parameter screening for oxygen
supply: first, the newly established principle of adjusting surface-to-volume
ratio that results in variation of oxygen transfer conditions within microtiter
plates. Second, the interface for transferring specific oxygen supply properties
between reactor systems, that facilitates to use screening data as a basis for
further process development. Both elements must be critically evaluated for
further application of the method.
As one influencing factor, potential side-effects of changing filling volumes in
BioLector® cultivation must be excluded for obtaining valid results. Due to the
consistency of the obtained data, it is assumed that all analytical side-effects
were identified and mitigated in the presented study. The main new
development is the multi-parameter calibration based on backscatter, volume
and biomass, which makes it possible to compare parallel cultivations of
32
different filling volume. This calibration is a vital requirement with respect to
determination of valid biomass specific data, e.g. yield or product formation at
different oxygen transfer capacity. Therefore, there was no alternative to its
implementation. The quality of the obtained biomass data can be assessed
based on the variances that were observed during the method characterization
and screenings. The degree of accuracy is in a range below 5% standard
deviation above 2 g·L-1 of cell dry weight for triplicate measurements, which
is acceptably low for comparative yield determination. Error rates are higher at
low biomass concentration and filling volume, which can be explained by
physical factors related to the measuring principle, e.g. increased surface
reflections at thinner liquid coverage of Flowerplate base foil. These
irregularities are factored in by the multi-parameter calibration, which
therefore forms a reliable enhancement of the BioLector® cultivation system.
It is vital for the screening method that specific oxygen-transfer results can be
used as a basis for process development in other reactor systems. However, the
initial study design was based on a sulfite-oxidation system19, for which
concerns about the precision have been raised20. Also, the sulfite oxidation
system belongs to the class of capacity-based systems for oxygen transfer
determination, which is prone to systematic errors due to its dependency on
reference experiments. Therefore, validation of the oxygen transfer properties
was performed with a balance-based system for oxygen transfer determination.
Balance-based systems, e.g. the respiration activity monitoring system
(RAMOS) and exhaust-gas analysis, are the most reliable option for process
transfers due to their online determination of actual oxygen transfer rates. In
particular, RAMOS has been proven as a reliable alternative to exhaust-gas
analysis of active C. glutamicum cultures21. Unfortunately, balance-based
systems are also hard to implement or miniaturize into systems of higher
cultivation throughput that are most suitable for screenings. The recently
introduced microtiter-compatible MTP-RAMOS system forms a compromise:
it can be applied for microtiter plates, although oxygen transfer can only be
determined for the whole microtiter plate22. In the presented study, the
capacity-based oxygen transfer control was applied for the screening, and then
33
verified with the respective RAMOS system. This combination provides the
most reasonable trade-off between high throughput and reliability of results.
This way, maintaining precise oxygen transfer control during scale-up to
subsequent process stages is feasible, and the strength of balance-based
determination of oxygen transfer can be accessed as a reference standard for
large-scale bioprocess development early on.
4.1.2 Evaluation of oxygen supply screening results
The biological properties of C. glutamicum that were identified during the
screening directly influenced the quality of the obtained characterization
dataset. It was demonstrated that C. glutamicum has a remarkably high
tolerance against very low levels of oxygen supply in its biomass and product
yields. The metabolic behavior during oxygen supply limitation was
successfully characterized, but the most relevant impacts on process yields
were observed at the lowest possible supply settings of the chosen study
design, i.e. below an oxygen transfer capacity of 8 mmol·L-1·h-1. This means
that most of the obtained results can be expressed as a robustness threshold,
while only a small fraction of the screened settings yielded deeper insights
beyond the robustness of the aerobic phenotype. The resolution of the
screening results could therefore be increased if respective means for stronger
oxygen supply limitation were taken. This could be achieved by a number of
design changes, e.g. increase of initial batch carbon source or decrease of
agitation. Due to the systematic approach of the screening, both means would
make it necessary to devise a new study and screening layout.
The high tolerance for oxygen supply limitation also calls more attention to the
behavior at anaerobic conditions. Analysis of metabolism without oxygen
supply was not in the focus of this study. The reason for this is that anaerobic
conditions mark an extreme point without considerable growth or maintenance
for the aerobic C. glutamicum. They have little relevance for bioprocesses that
rely on regeneration of biomass, which is the typical setting of amino acid
production, and were therefore neglected in the presented study. However, it is
34
generally possible to include anaerobic conditions into an oxygen supply
screening approach with anaerobic BioLector® technology, which has been
established recently23. Microaerobic conditions, on the other hand, have
successfully been analyzed at the lower oxygen transfer settings of the
presented screening data. Further adjustment of the process conditions could
be applied for reducing the oxygen transfer rate (see previous paragraph), and
could easily be screened with more resolution in a follow-up study.
As the screening has shown, there is a broad range within aerobic to partially
limited oxygen supply levels which neither influences the biomass nor product
yield. Quick yield determination based on oxygen transfer conditions is the
key strength of the screening method, so that the robustness threshold values
are the main result of the screening. However, dissolved oxygen profiles were
also determined for all wildtype cultivations as a supplement to the yield data.
Their main purpose was the verification of oxygen supply limitation, and
respective raw data were not presented in the manuscript. It was observed that
the dissolved oxygen measurements were more prone to technical
malfunctions than in other BioLector® setups, which may have been a result of
insufficiently covered base foil at agitated conditions and low filling volume.
However, even with this quality concern, correlation of a major fraction of the
dissolved oxygen data with growth and side product secretion was possible. It
can be considered as a positive quality attribute for the screening method that
in-process data are available for dissolved oxygen, so that calculation of
oxygen transfer rates is principally possible. However, the system may yet
require some modifications if quantitative data become necessary for future
studies.
As was summarized in the publication, the generated results do not contain
fundamentally new information about the metabolic pathways of C.
glutamicum, because its aerobic and anaerobic pathways have been
characterized extensively in the past8. The optimum of L-lysine productivity is
in the fully aerobic range, which confirms studies that were performed with
other producer strains11,24. An interesting element of the results is the fact that
35
growth rate is not directly affected when dissolved oxygen levels are fully
depleted, which can be deduced from the dissolved oxygen signals to the
respective growth curves. There seems to be a metabolic buffering potential,
which is connected to increased side product turnover rates. Only after this
intermediate phase has passed, there is a reduction in growth. This strongly
suggests that metabolic energy supply is in fact a very dynamic and flexible
system that facilitates sufficient output for anabolism even at changing
environmental conditions.
An intrinsic quality of all obtained screening results is the transferability into
other reactor systems (see 4.1.1), which is heavily influenced by the applied
organism and cultivation medium. Therefore, the presented study contains a
combination of a simple oxygen transfer algorithm based on sulfite oxidation
for screening design in the BioLector®, and a precise kinetic study in a
balance-based reactor system (RAMOS) for reference. Oxygen-limited growth
was confirmed by RAMOS with characteristic kinetics of exponential-plateau-
stationary succession in oxygen uptake rates. This combination provides the
ease and flexibility that is necessary to screen a parameter range together with
the sound characterization of oxygen transfer behavior over cultivation time.
The biological properties of any given microorganism make it indispensable to
put some effort into finding a suitable strategy for transfer between reactor
systems. In this case, the transfer constant determined with the RAMOS
system can serve as a gateway to other reactor systems that rely on balance-
based approaches to oxygen transfer monitoring.
It is the main strength of the oxygen transfer screening that the replicate extent
and resolution of the generated dataset on oxygen supply impact is unique in
the field of microbial bioprocess characterization. The ease of performing
parallel cultivations with a full range of applicable supply conditions cannot be
matched with stirred tank investigation. Alternative reactor systems with
higher throughput so far either lacked the potential for changing the oxygen
transfer rate from unit to unit, or could not be characterized sufficiently to
ensure relevance of specific conditions for larger reactors. Although the
36
presented screening is yet limited to a rather simple batch process, the
potential for incorporating more advanced reactor conditions increases with
recent progress in process miniaturization, which will be discussed in section
4.3.
As was presented in the Results section, several screened cultivation
conditions were subjected to a series of scale-up experiments to 5 liter stirred-
tank cultivation (data not shown in publication). This was intended as a quality
check and extension for the presented oxygen transfer screening data.
Qualitatively, these experiments provided similar results as the screening.
However, several challenges emphasized the importance of proper interface
design (i.e. scale-up criteria) when dealing with very low oxygen supply
conditions. As had previously turned out during the screening, the most
metabolically relevant range of oxygen supply lies in the region of oxygen
transfer capacity OTRmax < 8 mM*h-1, where growth and side product
accumulation characteristics are heavily dependent on oxygen supply. This
range was also subject to technical limitations in precise control of oxygen
supply in the stirred-tank setup, which is why the exact matching of the
oxygen transfer capacity in the BioLector® and the stirred tank was hard to
achieve. Both systems were in their lowest feasible range of gassing and, in the
case of the stirred tank, agitation, and secondary effects like decreased gas
exchange (stirred tank) may have had an influence on the respective
characterization data. Transferring high oxygen transfer conditions, on the
other hand, was entirely unproblematic, but also without phenotypic changes
being associated with the magnitude of oxygen supply. Therefore, selected
scale-up of screened oxygen-limited batch cultivations into 5 liter lab-scale
stirred tank provided only a limited set of useful data that generally confirmed
the screening results. Verification of highly oxygen-limited scale-up for
scientific purposes would require to apply the MTP-RAMOS reference, as has
been described above, which was not possible in the timeframe of the
presented study. However, this step could easily be included into future
experimental designs when the magnitude of suitable oxygen transfer settings
can be estimated with more precision.
37
4.2 Scale-Down of oxygen transfer inhomogeneity into lab-
scale
Scale-down of metabolic stress factors in inhomogeneous process
environments has been established over a long period of time and proceeds to
offer new insights into metabolic behavior under industrially relevant
conditions, as has been summarized in recent perspective publications25–27.
Paper II2 and Paper III3 are applications of an established inhomogeneity scale-
down technique for C. glutamicum: cultivation in two-compartment stirred
tank/ plug flow reactors (TCR or STR-PFR) for simulation of large-scale fed-
batch situations. They provide a systematic description of metabolic
robustness under inhomogeneous oxygen and substrate supply. In the
following sections, the results will be evaluated in their relevance for industrial
application, and in their quality and limitations compared to other
inhomogeneity scale-down sources.
4.2.1 Evaluation of relevance for industrial application
The study designs for this work are based on the assumption that parallel
oscillation of oxygen and substrate availability is a crucial influencing factor
for large-scale bioprocesses. This stress factor, among others, is widely
accepted in literature as one of the inhomogeneity-related causes for failure in
large-scale process development25–27. For organisms with high metabolic
activity, scaling complications of oxygen and substrate supply inhomogeneity
are especially likely to occur. The reason is that these substrates have very
high metabolic turnover in active bacterial cultures compared to characteristic
mixing times of large-scale reactors27. As a rule of thumb, oxygen depletion
can take seconds at substrate saturation, while gradient depletion through
mixing is in the minute range25. Fed-batch cultivation, which is applied in
most industrial processes with C. glutamicum, intensifies this challenge due to
the local addition of concentrated substrates. Of course, there are also other
factors besides oxygen and substrate supply which are subject to process
38
inhomogeneity in typical large-scale bioprocesses: for example temperature,
pressure, shear forces, or concentration of dissolved gasses. However, these
other inhomogeneities are typically less demanding for mixing power, and can
therefore be neglected in comparison to oxygen and substrate supply.
Additionally, many influencing factors of large-scale bioprocesses are not
related to inhomogeneity, e.g. culture times and media preparations. These
scaling influences that are not related to inhomogeneity can usually be
included into lab-scale assessment with small experimental effort, e.g. by
adjusting preculture conditions or using industrial substrates. Therefore, it can
be concluded that the assessed scale-down topic is highly relevant for
industrial purposes, and can be applied as a simplified robustness indicator for
simplified assessment of complex bioprocess scaling challenges.
Industrial relevance of inhomogeneity scale-down assessment is heavily
dependent on the choice of the applied reactor system, in this case the STR-
PFR. There are many alternative scale-down systems available which can be
applied for subjecting microbial cell cultures to oscillating environments26. For
example, mixing time increase can also be reached by disc installation inside a
stirred tank28 or STR-STR reactors29. The main advantage of STR-PFR
systems is that they can distinguish between time-dependent metabolism
during the course of oscillations, and overall inhomogeneity effects on an
inhomogeneous culture. The relatively narrow PFR residence time distribution
does not necessarily improve the simulation quality compared to other designs,
e.g. STR-STR, as will be discussed below. However, sampling the PFR part
of the reactor yields quantitative data about turnover of metabolites under step-
change conditions. This allows studying the reasons for metabolic robustness
or susceptibility against oscillations, rather than just providing information
about the degree of robustness. Therefore, the choice of STR-PFR as scale-
down reactor system is a compromise between the quality of metabolic
characterization and quality of large-scale inhomogeneity simulation.
The quality of prediction for the presented studies, i.e. whether the obtained
results could be reproduced under industrial process conditions, is restricted by
39
the lack of relevant industrial reference data. In order to provide precise
scalability results for a bioprocess, it is important that the design of a scale-
down reactor is based on a valid process model for the target large-scale
system. In principle, the STR-PFR with feed injection into the unaerated PFR
is designed to simulate the mixed bulk and feed zone of a top-fed industrial
bioreactor. Fed-batch processes are the typical setup for highly active bacterial
processes in large-scale, with C. glutamicum amino acid production being no
exception30. An example for a model-based scale-down study of this particular
process situation is detailed in Lapin et al.31, who applied mechanistic and
fluid-phase models to distinguish between reactor zones of high and low
substrate availability for an E. coli process. Theoretical background on scale-
down design has also been described by Delvigne et al.32, who used stochastic
modelling for STR-PFR dimensioning based on large scale mixing
characteristics. However, simulation quality of any scale-down study depends
on large-scale mixing characterization for existing or simulated production
plants in the target scale. This can be achieved by computational fluid
dynamics (CFD) models of a planned reactor system. Based on the availability
of these large-scale data, the adjustment and characterization of respective
mixing properties in the scale-down reactor is a precondition for adequate
simulation. Suitable procedures for mixing characterization of reactor
compartments with continuous circulation are outlined in Levenspiel33 for
calculating Bodenstein-numbers (Bo) in plug flow compartments based on the
axial dispersion model, which is highly relevant for STR-PFR setups. It is the
main drawback for the presented studies that there is currently no published
characterization for state-of-the-art industrial stirred tank reactors, which are
applied for amino acid production with C. glutamicum. Large-scale mixing
data for realistic reactor systems, which reach stirred tank volumes of ca.
500 m330 up to 750 m37, could also not be obtained from industrial partners due
to corporate confidentiality. Therefore, the presented studies were set up as a
worst-case approach for an estimated inhomogeneity assessment. Residence
times and volume proportions of the two compartment reactors were chosen in
a resembling range to previous studies with B. subtilis34, E. coli35 and
40
S. cerevisiae36 to facilitate robustness comparison between the biological
systems. Transferability of the results towards industrial application is
therefore limited to the qualitative description of metabolism at
inhomogeneous conditions, so that later simulations for target applications can
principally be matched with the available data. In spite of the lack of a specific
target reactor, the applied scale-down systems were fully characterized in their
mixing properties to facilitate their incorporation into suitable large-sale
simulations.
4.2.2 Evaluation of inhomogeneity scale-down results
The presented manuscripts provide extensive data about STR-PFR
inhomogeneity scale-down cultivation for C. glutamicum, with a sequential
assessment of general metabolic behavior in a wildtype strain2 and detailed
robustness-assessment for a producer strain of L-lysine3. The manuscripts
complement one another due to their shared inhomogeneity scale-down
approach, although strategy and focus are different except for the general
setup. In the following, contents of both manuscripts are summarized and
compared to literature.
The wildtype-paper2 comprises analysis of batch and linear feed conditions. It
is a broad approach for description of process inhomogeneity effects that
compares anaerobic circulation times of 45 and 87 seconds to a homogeneous
reference culture. The study was conducted as a basic assessment of
metabolism at inhomogeneous process conditions, and therefore contained a
limited set of analytics and cultivation replicates. This provided insight into
intracellular metabolism and side product circulation between the aerobic and
oxygen-limited zones of the STR-PFR. The study identified comparable
growth, similar pool sizes of adenosine phosphates and reducing equivalents,
and constant pH shifts during anaerobic holding time for the inhomogeneous
processes. A comparison of batch versus feed phase was made for oxygen
uptake redistribution, with the feed phase showing similar breathing activity
between STR-PFR and homogeneous reference cultivation. The batch phase
41
showed reduced breathing activity in the inhomogeneous system, and was
therefore assumed to be less robust against inhomogeneity. The main result of
the paper is the identification of the changed carbon metabolism with futile
cycling of side products into extracellular space and back, the resulting mixed-
substrate growth, and general information about robustness compared to other
organisms that were assessed in reactor systems with similar scale-down
settings.
The second scale-down study with the producer strain for L-lysine was partly
based on the robustness results of the wildtype study, because it contains an
emphasis on more analytical depth and increased process inhomogeneity. A
focus on extended exponential fed-batch phases was chosen for robustness
assessment, after initial analysis in the wildtype study had proven that oxygen
uptake redistribution only occurs in a substrate-limited environment, whereas
batch growth inhomogeneity leads to reduced breathing activity. Residence
times of up to 180 seconds were assessed in the oxygen-limited PFR. This
extreme exposure to the substrate excess/ oxygen limitation environment is in
the range of mixing times that can be expected in stirred tank reactor systems
of above 30,000 liters37, and exceeds the conditions that were applied in the
available inhomogeneity scale-down literature with comparable STR-PFR
systems. An exponential feed profile was applied, which generates more stable
substrate-limited process conditions over cultivation time than the linear feed
in the wildtype study. This improves sensitivity for the detection of changes in
growth parameters caused by inhomogeneity. Also, several cultivation
replicates and redundant reactors were applied for each condition which
improves precision and variance detection for the study results. Analytics of
intracellular adenosine phosphates and reducing equivalents yielded
comparable results to the wildtype study, even at highly increased
inhomogeneity, which is a strong indicator for metabolic robustness. Targeted
peptide quantification and maximum substrate uptake capacity were added to
the analytic spectrum as indicators for proteome and enzymatic robustness
against inhomogeneity. The L-lysine product yield was similarly unchanged
and robust against high degrees of inhomogeneity. The pH shift which was
42
observed for the wildtype also occurred with the producer strain, with the
extent of pH decrease correlating to biomass and feed profile. The futile
conversion of substrate into extracellular side products was also identified for
the producer strain, although to a smaller extent than in the wildtype (up to
16% carbon fraction turnover, cf.2).
The common conclusion of both manuscripts is that C. glutamicum possesses
high robustness against oscillating oxygen and substrate supply in a fed-batch
environment. Integration of multiple analytical methods provided more insight
into the nature of process inhomogeneity in oxygen and substrate supply.
Repeated exposure to oxygen supply limitation does not trigger any
uncommon regulation pathways, at least not in the assessed anaerobic
circulation times in the minute range. The phenomenon of side product
secretion and reabsorption has been analyzed in detail and can be assumed to
be a direct, yet often unrecognized side effect of process inhomogeneity for
C. glutamicum. The presented results are a strong indication that the
additionally required cross-membrane transport steps of the side product
lactate do not have a negative impact on metabolism and growth. The main
function of the reversible transport mechanism appears to be the maintenance
of fast substrate uptake and conversion at oxygen supply limitation, when
respiratory chain action is impaired. Under these conditions, NAD+ is
regenerated by the conversion of pyruvate into lactate through the action of
lactate dehydrogenase. Previously, several studies have focused on adaptation
processes and metabolic challenges in growth on lactate, which can now be
assessed against the present study results that imply mixed substrate growth in
inhomogeneous environments. Literature shows that the activity of several
central metabolic enzymes is regulated under conditions of growth on lactate:
phosphoenolpyruvate-synthethase is upregulated by a factor of two to three38,
glucose-6-phosphate-dehydrogenase and 6-phosphogluconate-dehydrogenase
are downregulated39–41, pyruvate carboxylase is upregulated by a factor of
three42,43, and phosphoenolpyruvate carboxykinase is upregulated by a factor
of two44. Furthermore, protein expression is changed under oxygen deprivation
conditions: for example, increased lactate dehydrogenase expression was
43
shown by Inui et al.45. Neither of these regulation effects could be identified in
the presented studies on the level of transcriptome or proteome, which
illustrates the metabolic robustness threshold of C. glutamicum against
inhomogeneity of oxygen and substrate supply. This is also a strong indication
for robustness against the mixed-substrate conditions of organic acid and
sugar, which are a direct consequence of inhomogeneous oxygen and substrate
supply. Furthermore, the number of cultivation replicates and the two analyzed
strains is a big advantage of the performed studies, because it provides an
increased security that the observed metabolic behavior is reproducible. This
quality attribute is important for the characterization of a robustness behavior,
which otherwise harbors the risk that small metabolic changes might be
overlooked as typical batch-to-batch variance.
The presented study results of Paper II2 and Paper III3 must be assessed against
the past and recent observations of other groups, who have chosen different
approaches for scale-down of industrial C. glutamicum mixing properties.
There are currently three peer-reviewed sources that describe comparable
studies in compartmented reactors. The first one by Schilling et al.28 applies
disc installations in a stirred tank with an auxotrophic L-lysine producer strain,
which induces mixing time increase in the minute range for postulated
simulation of a 10 m3 stirred tank cultivation. Growth and critical enzyme
activities are negatively affected by this degree of inhomogeneity. Due to the
strain’s auxotrophy, the focus of the study shifts from oxygen and carbon
source inhomogeneity to the distribution of the limiting amino acid. Therefore,
the fact that a negative impact of inhomogeneity is only observed in the case
of Schilling et al., and not in the presented studies, can most likely be
attributed to the special properties of the applied biological system. Due to
selection pressure for efficient stress response during the previous evolution of
C. glutamicum, non-auxotrophic strains are more robust against temporary
oxygen or substrate depletion, while an auxotrophic strain lacks adequate
regulatory responses for oscillating availability of the essential amino acid.
Furthermore, the reactor design which was applied by Schilling et al. creates a
complicated residence time distribution for individual cells within the
44
insufficiently mixed reactor zones. While reduced back-mixing in a STR-PFR
defines a narrow time range for exposure to harsh growth conditions,
Schilling’s reactor may have areas resembling dead zones that can capture
cells over extended timespans. Due to a lack of position-specific reactor
characterization data, these dead zones may generate unknown degrees of
oxygen supply limitation with availability of the limiting amino acid and
substrate supply, which could generate increased stress compared to non-
auxotrophic strains. Therefore, it is highly probable that Schilling’s
observation of decreased robustness against inhomogeneous cultivation was
caused by the biological and reactor type difference, and that a comparison to
the results of the presented studies is therefore not applicable.
Two more inhomogeneity scale-down papers were published recently after the
presented manuscripts, which utilize three compartment reactors: Lemoine et
al.46 apply defined substrate depletion/excess along with oxygen supply
limitation in two coupled PFRs, thus simulating distinct zones in a bottom-fed
industrial stirred tank. A producer strain of L-lysine was used for cultivation.
Lemoine observes similarities in the cellular energy charge that resemble
results of the presented manuscripts, but also reports enhanced organic acid
accumulation and reduced product yield in scale-down cultivations. This result
is surprising because the extent of anaerobic residence time applied in
Lemoine is lower than in one of the presented manuscripts3. Also, this effect is
similarly observed without substrate depletion, i.e. if only one PFR is applied.
Due to the similarity of the setup to the presented studies, this different
observation can most likely be attributed to subtle differences in the biological
system. The observed metabolite pool and nucleotide data support that
C. glutamicum DM1800 might well be more susceptible to process
inhomogeneity than DM1933. On the other hand, Buchholz et al.47 apply an
STR cascade of dissolved gas gradients concerning CO2/HCO3- with
industrially relevant residence times in the minute range. Buchholz reports
overall process robustness against inhomogeneity of CO2, although several
responses concerning transcription are identified. This complements the
findings of the presented studies in the sense that quantifiable perturbations in
45
CO2 levels are not a concern for process scale robustness, and thus confirms
the assumptions made in section 2.3.2.
A comparison with inhomogeneity scale-down analysis of other biological
systems points out the uniqueness of the identified robustness against process
inhomogeneity in oxygen and substrate supply. Starting from early analysis of
S. cerevisiae36 and P. chrysogenum48, many industrial organisms and strains
have been subjected to respective scale-down studies. Perhaps the best
comparison to the presented studies can be drawn with the extensive E. coli
characterization data, many of which have been summed up by Enfors et al.49.
From differences in transcription50 to substrate uptake capacity29 and side
product accumulation /reabsorption with reduced biomass yield51, various
changes in metabolism have been described in response to oxygen and
substrate supply inhomogeneity with the help of scale-down reactors. The
overall effect of process inhomogeneity on bioprocesses is detrimental to
product and biomass yield, with very few exceptions (e.g. increased E. coli
cell viability at inhomogeneous process conditions reported by Hewitt et al.52).
Judging from these results, C. glutamicum can be considered as a positive
example of robustness against inhomogeneous process conditions in oxygen
and substrate supply. This means that in order to generate visible changes in
process outcome, harsh process conditions are required, which are beyond the
extent of large-scale stirred tank mixing times. In the presented studies, a
reasonable maximum of holding times in the range of previous studies31,34,36,49
has been applied. The fact that the metabolic robustness of C. glutamicum
exceeds these challenging environments means that there may be an unused
design space for cultivation under inhomogeneous conditions of oxygen and
substrate availability in C. glutamicum. This constitutes a positive example for
passing the scale-down test of a specific worst-case challenge.
46
4.3 Conclusion & outlook
4.3.1 Oxygen supply screening
Development of the screening method for metabolic impact of oxygen transfer
conditions was successful. The organism C. glutamicum is very robust against
oxygen transfer limitations, which led to an unexpectedly intense stress test for
the screening method: the detected changes in biomass and product yield of
the presented study were close to the technically feasible limit of adjustable
oxygen transfer conditions. Side product detection has shown that data about
metabolic states during oxygen supply limitation can be generated in parallel
with a performance screening. Therefore, the new screening tool can replace
stirred tank investigation of bioprocess susceptibility against oxygen supply
limitation with the more powerful parallel assessment in microtiter plate
format. The gained systematic knowledge about incremental changes on a
model process provides a new dimension to the current metabolic
understanding of oxygen transfer limitation. The resulting database contains
individually monitored cultivations that cover the full range from aerobic to
highly oxygen-limited growth. It can be used to verify hypotheses and
determine the set points at which physical or chemical triggers change
metabolic flow profiles. The resolution of the screening system is sufficient for
identification of metabolic turning points and cause-and-effect chains through
mapping of all applicable supply situations. The methodology is ready for
implementation into mini-plant process development workflows.
It has to be considered that the presented approach for screening oxygen
supply as a process parameter requires more experimental effort than most
established screening applications, e.g. for strains or media composition. This
mainly corresponds to the need to establish a multi-parameter calibration and
system transfer reference for validation of oxygen transfer over cultivation
time. Thus, the initial preparation time for the screening is quite long.
However, the presented approach provides unprecedented speed and replicate
availability once the system has been established. Furthermore, the integration
47
of oxygen supply as another complex process parameter is a big improvement
for bioprocess miniaturization. Although the presented study was a batch
growth setup, there are autofeed alternatives53 and microfluidic solutions54 on
the rise. Also, screening for oxygen transfer demands fits into place with
several current developments in the field of bioprocess miniaturization. There
are many successful application examples of the BioLector®, bioREACTOR®
48 and other miniaturized, yet highly controlled parallel reactor systems.
Implementation of sophisticated process chains into the milli- to microliter
scale has become a viable alternative to the previous approaches with untested
strains and conditions, which have led to many development failures in the
past. From a Design of Experiments (DoE) perspective, every new parameter
that can be investigated in early phases means less time and resource demand
during the later stages of process development. Thus, a screening for the
effects of variations in oxygen transfer on the metabolism of prokaryotic cell
cultures was long overdue. Implementation into fully miniaturized process
development chains, e.g. enhanced BioLector® setups55, can improve the
quality of results from the initial process design phase to new dimensions.
4.3.2 Scale-down of process inhomogeneity
Application of the STR-PFR scale-down method for analysis of
inhomogeneous oxygen and substrate supply was successfully performed in a
worst-case assessment of metabolic robustness. This has provided versatile
new data to previously available studies of inhomogeneity scale-down for
C. glutamicum. The combination of state-of-the-art analytical methods with
established bioprocess systems has provided a comprehensive picture of
metabolic robustness against inhomogeneity. Increasing the anaerobic holding
times beyond usually analyzed extents has shown that C. glutamicum has
remarkable traits of flexibility in changing microenvironments, which is
marked by oscillation of metabolic flows. The combined analysis of short-time
reversible effects and robustness assessment through process yields has
provided a step into the direction of systems biology for inhomogeneous
48
cultures, and a deeper understanding and characterization of process dynamics
in an industrial context. Most pronouncedly, the scale-down experiments have
provided new information concerning the extent of dynamic redistribution of
carbon flow in inhomogeneous cultures with the subsequent formation of local
sub-populations that differ in their metabolic activity.
It can be concluded that adjustments for improved inhomogeneity robustness
on the level of metabolic engineering are not required for current production
strains of C. glutamicum.
4.3.3 Outlook
This work is an appeal for early consideration of oxygen supply requirements
throughout all phases of bioprocess development in an industrial context. The
newly developed oxygen transfer screening and application of the
inhomogeneity scale-down can be applied in a comprehensive process
development chain. This could constitute a more efficient methodology for
aerobic bioprocess development, whenever the dependency of process-
outcome on oxygen supply is critical. The proposed workflow can be depicted
as follows (see Figure 4):
49
Figure 4: workflow comprehensive consideration of oxygen supply requirements during
bioprocess development, OTRmax = maximum oxygen transfer capacity, Default OTRmax =
arbitrary constant supply condition during screening, Target OTRmax = realistic supply
condition which is feasible for commercial scale
In the proposed development workflow, all oxygen supply elements of
bioprocesses are optimized in early development phases, and later transferred
into industrial scale application. This serves to prevent avoidable performance
loss. The critical step is that respective target conditions must be transferred
from the industrial reality to the respective process development phase at
which they become relevant. This means in particular that strain selection in
oxygen supply screening must be based on the achievable oxygen transfer
rates in industrial-scale reactors. Inhomogeneity scale-down for robustness
assessment must be the direct consequence once suitable strains have been
identified. This perspective on bioprocess development transfers the most
crucial development challenges into the reactor systems that can most easily be
operated, and thereby minimizes the possible sources of error during the cost-
intensive final development steps. The workflow is particularly well-suited for
50
implementation into a DoE- based assessment of new producer strain
candidates, with oxygen supply being one of the influencing factors for
performance indicators during screening, and inhomogeneity being a worst-
case or qualitative assessment parameter during lab-scale assessment.
Resources could be saved both due to more efficient resulting processes and
less futile process development steps for failing process transfers. Speed and
high resolution during screening, as well as deeper process understanding of
inhomogeneity is the key to avoid unnecessary scale-up failures due to
neglected oxygen supply demands.
In conclusion, industrial bioprocess development could benefit from
comprehensive consideration of oxygen supply requirements. Necessary
elements of a proposed workflow have been illustrated for the example of
C. glutamicum bioprocesses in this thesis. By implementing the presented
workflow, resulting bioprocesses could become more competitive during their
development process. This strategy could be applied for the benefit of aerobic
bioprocess development in general, thus mitigating one key complexity in
bioprocess engineering.
51
5 Literature
1. Käß, F. et al. Rapid assessment of oxygen transfer impact for Corynebacterium
glutamicum, Bioprocess Biosyst Eng 37, 2567–2577 (2014).
2. Käß, F. Junne, S. Neubauer, P. Wiechert, W. & Oldiges, M. Process inhomogeneity leads
to rapid side product turnover in cultivation of Corynebacterium glutamicum, Microb. Cell
Fact. 13, 6 (2014).
3. Käß, F. et al. Assessment of robustness against dissolved oxygen/substrate oscillations for
C. glutamicum DM1933 in two-compartment bioreactor, Bioprocess Biosyst Eng 37,
1151–1162 (2014).
4. Bauer, K. H. Lippold, B. C. & Breitkreutz, J. Pharmazeutische Technologie. Mit einer
Einführung in die Biopharmazie. 9th ed. (Wissenschaftliche Verlagsgesellschaft, Stuttgart,
2012).
5. Levin, M. Pharmaceutical process scale-up (Marcel Dekker, New York, ©2002).
6. Binder, S. et al. A high-throughput approach to identify genomic variants of bacterial
metabolite producers at the single-cell level, Genome Biol. 13, R40 (2012).
7. Eggeling, L. & Bott, M. Handbook of corynebacterium glutamicum (Taylor & Francis,
Boca Raton, 2005).
8. Bott, M. & Niebisch, A. The respiratory chain of Corynebacterium glutamicum, J.
Biotechnol. 104, 129–153 (2003).
9. Buschke, N. Schäfer, R. Becker, J. & Wittmann, C. Metabolic engineering of industrial
platform microorganisms for biorefinery applications – Optimization of substrate
spectrum and process robustness by rational and evolutive strategies, Bioresource
Technology 135, 544–554 (2013).
10. Ensari, S. & Lim, H. C. Kinetics of l-lysine fermentation: a continuous culture model
incorporating oxygen uptake rate, Applied Microbiology and Biotechnology 62, 35–40
(2003).
11. Guillouet, S. & Engasser, J. M. Growth of Corynebacterium glutamicum in glucose-
limited continuous cultures under high osmotic pressure. Influence of growth rate on the
intracellular accumulation of proline, glutamate and trehalose, Appl Microbiol Biotechnol
44, 496–500 (1995).
12. Follmann, M. et al. Functional genomics of pH homeostasis in Corynebacterium
glutamicum revealed novel links between pH response, oxidative stress, iron homeostasis
and methionine synthesis, BMC Genomics 10, 621 (2009).
13. Blombach, B. Buchholz, J. Busche, T. Kalinowski, J. & Takors, R. Impact of different
CO2/HCO3− levels on metabolism and regulation in Corynebacterium glutamicum,
Journal of Biotechnology 168, 331–340 (2013).
14. Eggeling, L. & Bott, M. A giant market and a powerful metabolism: L-lysine provided by
Corynebacterium glutamicum, Appl. Microbiol. Biotechnol. 99, 3387–3394 (2015).
52
15. Kinoshita, S. Udaka, S. & Shimono, M. Studies on the amino acid fermentation. Part 1.
Production of L-glutamic acid by various microorganisms, J. Gen. Appl. Microbiol. 50,
331–343 (2004).
16. UDAKA, S. Screening method for microorganisms accumulating metabolites and its use
in the isolation of Micrococcus glutamicus, J. Bacteriol. 79, 754–755 (1960).
17. Blombach, B. Hans, S. Bathe, B. & Eikmanns, B. J. Acetohydroxyacid synthase, a novel
target for improvement of L-lysine production by Corynebacterium glutamicum, Appl.
Environ. Microbiol. 75, 419–427 (2009).
18. Neuner, A. et al. Production of l-lysine on different silage juices using genetically
engineered Corynebacterium glutamicum, Journal of Biotechnology 163, 217–224 (2013).
19. Hermann, R. Walther, N. Maier, U. & Büchs, J. Optical method for the determination of
the oxygen-transfer capacity of small bioreactors based on sulfite oxidation, Biotechnol.
Bioeng. 74, 355–363 (2001).
20. Linek, V. Kordač, M. & Moucha, T. Evaluation of the optical sulfite oxidation method for
the determination of the interfacial mass transfer area in small-scale bioreactors,
Biochemical Engineering Journal 27, 264–268 (2006).
21. Seletzky, J. M. et al. An experimental comparison of respiration measuring techniques in
fermenters and shake flasks: exhaust gas analyzer vs. RAMOS device vs. respirometer, J.
Ind. Microbiol. Biotechnol. 34, 123–130 (2007).
22. Wewetzer, S. J. et al. Parallel use of shake flask and microtiter plate online measuring
devices (RAMOS and BioLector) reduces the number of experiments in laboratory-scale
stirred tank bioreactors, J Biol Eng 9, 2 (2015).
23. Kensy, F. Born, O. Otte, B. & Jennewein, S. Anaerobic High-Throughput Fermentation.
New Microscale Technology with Online Monitoring Facilitates Processes for Biofuel
Applications, Genetic Engineering & Biotechnology News (GEN) 31 (2011).
24. La Toma, M. K. Shvinka, I. Ruklisha, M. P. Sakse, A. K. & La Baburin. Cyanide-resistant
oxygen consumption of the lysine-synthesizing bacteria Brevibacterium flavum 22 LD,
Prikladnaia Biokhimiia i Mikrobiologiia 20, 95–100 (1984).
25. Takors, R. Scale-up of microbial processes: impacts, tools and open questions, J.
Biotechnol. 160, 3–9 (2012).
26. Neubauer, P. & Junne, S. Scale-down simulators for metabolic analysis of large-scale
bioprocesses, Curr. Opin. Biotechnol. 21, 114–121 (2010).
27. Lara, A. R. Galindo, E. Ramírez, O. T. & Palomares, L. A. Living With Heterogeneities in
Bioreactors: Understanding the Effects of Environmental Gradients on Cells, MB 34, 355–
382 (2006).
28. Schilling, Pfefferle, Bachmann, Leuchtenberger & Deckwer. A special reactor design for
investigations of mixing time effects in a scaled-down industrial L-lysine fed-batch
fermentation process, Biotechnol. Bioeng. 64, 599–606 (1999).
29. Sandoval-Basurto, E. A. Gosset, G. Bolivar, F. & Ramirez, O. T. Culture ofEscherichia
coli under dissolved oxygen gradients simulated in a two-compartment scale-down
system: Metabolic response and production of recombinant protein, Biotechnol. Bioeng.
89, 453–463 (2005).
53
30. Hermann, T. Industrial production of amino acids by coryneform bacteria, Journal of
Biotechnology 104, 155–172 (2003).
31. Lapin, A. Schmid, J. & Reuss, M. Modeling the dynamics of E. coli populations in the
three-dimensional turbulent field of a stirred-tank bioreactor—A structured–segregated
approach, Chemical Engineering Science 61, 4783–4797 (2006).
32. Delvigne, F. Destain, J. & Thonart, P. A methodology for the design of scale-down
bioreactors by the use of mixing and circulation stochastic models, Biochemical
Engineering Journal 28, 256–268 (2006).
33. Levenspiel, O. Chemical reaction engineering. 2nd ed. (Wiley, New York, 1972).
34. Junne, S. Klingner, A. Kabisch, J. Schweder, T. & Neubauer, P. A two-compartment
bioreactor system made of commercial parts for bioprocess scale-down studies: impact of
oscillations on Bacillus subtilis fed-batch cultivations, Biotechnol J 6, 1009–1017 (2011).
35. Neubauer, P. Haggstrom, L. & Enfors, S.-O. Influence of substrate oscillations on acetate
formation and growth yield inEscherichia coli glucose limited fed-batch cultivations,
Biotechnol. Bioeng. 47, 139–146 (1995).
36. George, S. Larsson, G. & Enfors, S.-O. A scale-down two-compartment reactor with
controlled substrate oscillations: Metabolic response of Saccharomyces cerevisiae,
Bioprocess and Biosystems Engineering, 249–257 (1993).
37. Vrábel, P. van der Lans, R. G. Luyben, K. C. Boon, L. & Nienow, A. W. Mixing in large-
scale vessels stirred with multiple radial or radial and axial up-pumping impellers:
modelling and measurements, Chemical Engineering Science 55, 5881–5896 (2000).
38. Jetten, M. S. M. Pitoc, G. A. Follettie, M. T. & Sinskey, A. J. Regulation of
phospho(enol)-pyruvate-and oxaloacetate-converting enzymes in Corynebacterium
glutamicum, Appl Microbiol Biotechnol 41, 47–52 (1994).
39. Bianchi, D. Bertrand, O. Haupt, K. & Coello, N. Effect of gluconic acid as a secondary
carbon source on non-growing L-lysine producers cells of Corynebacterium glutamicum.
Purification and properties of 6-phosphogluconate dehydrogenase, Enzyme and Microbial
Technology 28, 754–759 (2001).
40. Sugimoto, S. & Shiio, I. Regulation of 6-phosphogluconate dehydrogenase in
Brevibacterium flavum, Agric. Biol. Chem. 51:1257 (1987).
41. Sugimoto, S. & Shiio, I. Regulation of glucose-6-phosphate dehydrogenase in
Brevibacterium flavum, Agric. Biol. Chem. 51:101 (1987).
42. Koffas, M. A. G. Jung, G. Y. Aon, J. C. & Stephanopoulos, G. Effect of Pyruvate
Carboxylase Overexpression on the Physiology of Corynebacterium glutamicum, Applied
and Environmental Microbiology 68, 5422–5428 (2002).
43. Peters-Wendisch, P. G. et al. Pyruvate carboxylase from Corynebacterium glutamicum:
characterization, expression and inactivation of the pyc gene, Microbiology 144, 915–927
(1998).
44. Gerstmeir, R. et al. Acetate metabolism and its regulation in Corynebacterium
glutamicum, Journal of Biotechnology 104, 99–122 (2003).
54
45. Inui, M. et al. Metabolic Engineering of Corynebacterium glutamicum for Fuel Ethanol
Production under Oxygen-Deprivation Conditions, J Mol Microbiol Biotechnol 8, 243–
254 (2004).
46. Lemoine, A. Maya Martίnez-Iturralde, N. Spann, R. Neubauer, P. & Junne, S. Response of
Corynebacterium glutamicum exposed to oscillating cultivation conditions in a two- and a
novel three-compartment scale-down bioreactor, Biotechnol. Bioeng. 112, 1220–1231
(2015).
47. Buchholz, J. et al. CO2/HCO3 − perturbations of simulated large scale gradients in a
scale-down device cause fast transcriptional responses in Corynebacterium glutamicum,
Appl Microbiol Biotechnol 98, 8563–8572 (2014).
48. Larsson, G. & Enfors, S.-O. Studies of insufficient mixing in bioreactors: Effects of
limiting oxygen concentrations and short term oxygen starvation on Penicillium
chrysogenum, Bioprocess and Biosystems Engineering 3, 123–127 (1988).
49. Enfors, S. O. et al. Physiological responses to mixing in large scale bioreactors, J.
Biotechnol. 85, 175–185 (2001).
50. Schweder, T. et al. Monitoring of genes that respond to process-related stress in large-
scale bioprocesses, Biotechnol. Bioeng. 65, 151–159 (1999).
51. Xu, B. Jahic, M. Blomsten, G. & Enfors, S.-O. Glucose overflow metabolism and mixed-
acid fermentation in aerobic large-scale fed-batch processes with Escherichia coli, Applied
Microbiology and Biotechnology 51, 564–571 (1999).
52. Hewitt, C. J. Nebe-Von Caron, G. Axelsson, B. McFarlane, C. M. & Nienow, A. W.
Studies related to the scale-up of high-cell-densityE. coli fed-batch fermentations using
multiparameter flow cytometry: Effect of a changing microenvironment with respect to
glucose and dissolved oxygen concentration, Biotechnol. Bioeng. 70, 381–390 (2000).
53. Panula-Perälä, J. et al. Enzyme controlled glucose auto-delivery for high cell density
cultivations in microplates and shake flasks, Microb Cell Fact 7, 31 (2008).
54. Funke, M. et al. Microfluidic biolector-microfluidic bioprocess control in microtiter
plates, Biotechnol. Bioeng. 107, 497–505 (2010).
55. Unthan, S. Radek, A. Wiechert, W. Oldiges, M. & Noack, S. Bioprocess automation on a
Mini Pilot Plant enables fast quantitative microbial phenotyping, Microb. Cell Fact. 14, 32
(2015).
